Merge branch 'develop' into fix_vlog

Author: velconia

cmake/external/mkldnn.cmake

@@ -45,7 +45,7 @@ IF(${CBLAS_PROVIDER} STREQUAL "MKLML")
 ELSE()
     MESSAGE(FATAL_ERROR "Should enable MKLML when build MKLDNN")
 ENDIF()
-SET(MKLDNN_FLAG "-Wno-error=strict-overflow -Wno-error=unused-result")
+SET(MKLDNN_FLAG "-Wno-error=strict-overflow -Wno-error=unused-result -Wno-error=array-bounds")
 SET(MKLDNN_FLAG "${MKLDNN_FLAG} -Wno-unused-result -Wno-unused-value")
 SET(MKLDNN_CFLAG "${CMAKE_C_FLAGS} ${MKLDNN_FLAG}")
 SET(MKLDNN_CXXFLAG "${CMAKE_CXX_FLAGS} ${MKLDNN_FLAG}")
@@ -54,7 +54,7 @@ ExternalProject_Add(
     ${EXTERNAL_PROJECT_LOG_ARGS}
     DEPENDS             ${MKLDNN_DEPENDS}
     GIT_REPOSITORY      "https://github.com/01org/mkl-dnn.git"
-    GIT_TAG             "64e03a1939e0d526aa8e9f2e3f7dc0ad8d372944"
+    GIT_TAG             "21fb5f2af1dd14e132af4f1b79160977ee487818"
     PREFIX              ${MKLDNN_SOURCES_DIR}
     UPDATE_COMMAND      ""
     CMAKE_ARGS          -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}

paddle/fluid/API.spec

@@ -118,9 +118,10 @@ paddle.fluid.layers.label_smooth ArgSpec(args=['label', 'prior_dist', 'epsilon',
 paddle.fluid.layers.roi_pool ArgSpec(args=['input', 'rois', 'pooled_height', 'pooled_width', 'spatial_scale'], varargs=None, keywords=None, defaults=(1, 1, 1.0))
 paddle.fluid.layers.roi_align ArgSpec(args=['input', 'rois', 'pooled_height', 'pooled_width', 'spatial_scale', 'sampling_ratio', 'name'], varargs=None, keywords=None, defaults=(1, 1, 1.0, -1, None))
 paddle.fluid.layers.dice_loss ArgSpec(args=['input', 'label', 'epsilon'], varargs=None, keywords=None, defaults=(1e-05,))
-paddle.fluid.layers.image_resize ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'resample'], varargs=None, keywords=None, defaults=(None, None, None, 'BILINEAR'))
+paddle.fluid.layers.image_resize ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'resample', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, 'BILINEAR', None))
 paddle.fluid.layers.image_resize_short ArgSpec(args=['input', 'out_short_len', 'resample'], varargs=None, keywords=None, defaults=('BILINEAR',))
-paddle.fluid.layers.resize_bilinear ArgSpec(args=['input', 'out_shape', 'scale', 'name'], varargs=None, keywords=None, defaults=(None, None, None))
+paddle.fluid.layers.resize_bilinear ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, None))
+paddle.fluid.layers.resize_nearest ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, None))
 paddle.fluid.layers.gather ArgSpec(args=['input', 'index'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.scatter ArgSpec(args=['input', 'index', 'updates', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.sequence_scatter ArgSpec(args=['input', 'index', 'updates', 'name'], varargs=None, keywords=None, defaults=(None,))

paddle/fluid/inference/analysis/analyzer.cc

@@ -101,6 +101,7 @@ Analyzer::Analyzer() { Register("manager1", new DfgPassManagerImpl); }
 
 void Analyzer::Run(Argument* argument) {
   std::vector<std::string> passes;
+  passes.push_back("graph_viz_pass");  // add graphviz for debug.
 #ifdef PADDLE_WITH_MKLDNN
   if (use_mkldnn_) {
     VLOG(30) << "Adding MKL-DNN placement pass";
@@ -110,13 +111,13 @@ void Analyzer::Run(Argument* argument) {
   // infer_clean_graph_pass should be the first default pass
   // after mkldnn_placement_pass.
   passes.push_back("infer_clean_graph_pass");
+  passes.push_back("graph_viz_pass");  // add graphviz for debug.
   for (auto& pass : ir_passes_) {
     if (!disabled_ir_passes_.count(pass)) {
       passes.push_back(pass);
       passes.push_back("graph_viz_pass");  // add graphviz for debug.
     }
   }
-  passes.push_back("graph_viz_pass");
   argument->Set(kFluidToIrPassesAttr, new std::vector<std::string>(passes));
 
   for (auto& x : data_) {

paddle/fluid/inference/api/analysis_predictor.h

@@ -13,6 +13,8 @@
 // limitations under the License.
 
 #pragma once
+#include <algorithm>
+#include <map>
 #include <string>
 #include <vector>
 #include "paddle/fluid/framework/naive_executor.h"

paddle/fluid/inference/io.cc

@@ -59,7 +59,8 @@ void ReadBinaryFile(const std::string& filename, std::string* contents) {
 bool IsPersistable(const framework::VarDesc* var) {
   if (var->Persistable() &&
       var->GetType() != framework::proto::VarType::FEED_MINIBATCH &&
-      var->GetType() != framework::proto::VarType::FETCH_LIST) {
+      var->GetType() != framework::proto::VarType::FETCH_LIST &&
+      var->GetType() != framework::proto::VarType::RAW) {
     return true;
   }
   return false;

paddle/fluid/inference/tensorrt/engine.h

@@ -134,7 +134,7 @@ class TensorRTEngine : public EngineBase {
   std::unordered_map<std::string /*name*/, std::unique_ptr<framework::Tensor>>
       weight_map;
 
-  // TODO: (NHZLX)
+  // TODO(NHZLX)
   // In the normal case, the paddle-trt exists bug when runing the googlenet.
   // When there are more than two convolutions of 1 * 1 with the same input, the
   // paddle-tensorrt will do the merging optimization, which fuse those conv

paddle/fluid/operators/add_position_encoding_op.h

@@ -66,9 +66,10 @@ class AddPositionEncodingKernel : public framework::OpKernel<T> {
           x_lod.empty() ? max_seq_len : x_lod[0][i + 1] - x_lod[0][i];
       for (int j = 0; j < max_length; ++j) {
         for (int k = 0; k < half_size; ++k) {
-          const double val = (half_size > 1)
-                                 ? j / pow(10000.0, double(k) / (half_size - 1))
-                                 : j / 10000.0;
+          const double val =
+              (half_size > 1)
+                  ? j / pow(10000.0, static_cast<double>(k) / (half_size - 1))
+                  : j / 10000.0;
           dst_ptr[k] = src_ptr[k] * alpha + sin(val) * beta;
           dst_ptr[half_size + k] =
               src_ptr[half_size + k] * alpha + cos(val) * beta;